IF a random sample of unrelated individuals can be classified into genotypes for a particular genetic factor, a simple count of genes leads to efficient estimates of the relative frequencies of the two or more allelomorphs. If dominance makes certain genotypes phenotypically indistinguishable, the estimation of gene frequencies may be more troublesome, though, once the dominance relationships are understood, a method can readily be devised by application of the principle of maximum likelihood ; for a pair of allelomorphs, this amounts to no more than a count of recessives, but when several allelomorphic genes are involved, as for the ABO blood groups (Stevens, 1938), greater complexities arise. Often, however, the members of a sample from which gene frequencies are to be estimated are not wholly unrelated. Fisher (194o) has pointed out that the simple method of estimation just described is still consistent, but that it will in general be less precise than if it were based on an "unrelated" sample. If an assessment of precision is required, one of three procedures must then be followed: the sample might be reduced by rejection of all but one from each related group, so as to enable the estimation to be made by the old method ; a theoretical investigation of the precision of the estimate obtained by treating the sample as though its members were unrelated might be undertaken or an entirely new method of estimation, with its own assessment of precision, might be developed as particularly appropriate to the types of relationship encountered. The first of these alternatives is undesirable, since the choice of individuals to be rejected will introduce ambiguity and, in general, the rejects will be capable of giving additional information on the gene frequencies. Fisher considered the third alternative in its application to the particular problem of a simple recessive gene he developed maximum likelihood scoring systems for pairs of individuals consisting of parent and child, sibs, or haif-sibs. The method is analagous to the systems of scoring for detection of genetic linkage discussed by Fisher (i a and b ; 1946) and Finney (1940, 1941, 1943). Unfortunately, extension of the scoring technique to

中文翻译：

---

从家族记录估计基因频率

如果可以将无关个体的随机样本分类为特定遗传因素的基因型，那么简单的基因计数可以有效估计两个或多个等位基因的相对频率。如果优势使某些基因型在表型上无法区分，则基因频率的估计可能会更加麻烦，但是，一旦了解了优势关系，就可以通过应用最大似然原理轻松设计出一种方法；对于一对等位基因，这相当于隐性基因的数量，但是当涉及多个等位基因时，如 ABO 血型 (Stevens, 1938)，会出现更大的复杂性。然而，通常用于估计基因频率的样本成员并非完全不相关。Fisher (194o) 指出刚才描述的简单估计方法仍然是一致的，但它通常不如基于“无关”样本的精确。如果需要对精确度进行评估，则必须遵循以下三个程序之一：通过拒绝每个相关组中除一个之外的所有样本来减少样本，以便能够通过旧方法进行估计；可以对通过将样本视为不相关的样本进行处理而获得的估计精度进行理论研究，或者可能会开发一种全新的估计方法，具有自己对精度的评估，特别适合所遇到的关系类型. 这些替代方案中的第一个是不可取的，因为选择被拒绝的个体会引入歧义，而且一般来说，被拒绝者将能够提供关于基因频率的额外信息。费舍尔考虑将第三种替代方法应用于简单隐性基因的特定问题，他为由父母和子女、同胞或 haif-sibs 组成的成对个体开发了最大似然评分系统。该方法类似于 Fisher (ia and b ; 1946) 和 Finney (1940, 1941, 1943) 讨论的遗传连锁检测评分系统。不幸的是，将评分技术扩展到 费舍尔考虑将第三种替代方法应用于简单隐性基因的特定问题，他为由父母和子女、同胞或 haif-sibs 组成的成对个体开发了最大似然评分系统。该方法类似于 Fisher (ia and b ; 1946) 和 Finney (1940, 1941, 1943) 讨论的遗传连锁检测评分系统。不幸的是，将评分技术扩展到 费舍尔考虑将第三种替代方法应用于简单隐性基因的特定问题，他为由父母和子女、同胞或 haif-sibs 组成的成对个体开发了最大似然评分系统。该方法类似于 Fisher (ia and b ; 1946) 和 Finney (1940, 1941, 1943) 讨论的遗传连锁检测评分系统。不幸的是，将评分技术扩展到